Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces VLAGuard, a framework designed to evaluate and mitigate the vulnerability of Vision-Language-Action (VLA) robots to physical attention hijacking in wireless sensor networks. The authors develop a stress-test module called Visuomotor Attention-guided Semantic Attack (VASA) to simulate adversarial distractions and propose Attention-Protective Fine-Tuning (APFT) as a defense mechanism that stabilizes attention pathways without incurring inference overhead. Experimental results show that APFT significantly reduces failure rates in simulated environments and improves success rates in real-world trials, underscoring the importance of safeguarding attention mechanisms in VLA robots.
APFT slashes the failure rate of VLA robots from 100% to 25.9% against adversarial attacks, showcasing a breakthrough in attention stability.
Deploying Vision-Language-Action (VLA) robots as mobile edge nodes within wireless sensor networks (WSNs) requires robust protection against physical adversarial threats. We present VLAGuard, a framework to assess and mitigate a critical vulnerability: policy-critical action-to-vision attention hijacking. We first introduce a stress-test module, Visuomotor Attention-guided Semantic Attack (VASA), using printable patches to severely distract the robot's action-conditioned cross-attention. To counter this, we propose Attention-Protective Fine-Tuning (APFT), a defense that stabilizes spatiotemporal attention and enforces geometric consistency with zero inference overhead. Evaluations across simulated and physical WSN-assisted smart environments demonstrate significant robustness gains. APFT reduces the OpenVLA failure rate from 100.0% to 25.9% in LIBERO simulations. Furthermore, across 2,000 real-world trials, APFT improves the average success rate from 23.0% to 67.4% under severe patch attacks. This highlights that protecting attention pathways is important for improving the robustness of VLA-driven edge nodes in sensor networks.